A Nomogram for stratifying the malignancy risk for BI- RADS 4 breast masses detected on supplemental ultrasound in dense breast women

Abstract

Abstract Supplemental ultrasound is an effective way to increase the sensitivity of screening mammography for detecting breast cancer in women with dense breasts. However, due to its low positive predictive value (PPV), it often results in numerous unnecessary biopsies. This study aims to develop a predictive model that can stratify the malignancy risk of BI-RADS category 4 breast masses, which are identified additionally through supplemental ultrasound after screening mammography in women with dense breasts. After applying inclusion/exclusion procedures, a total of 425 eligible masses were selected from our institutional medical database. These masses were then divided into a training set (n=298) for model construction and a validation set (n=127) for model validation. A logistic regression model including five predictive characteristics was constructed and a corresponding nomogram was generated. The predictive model demonstrates robust calibration, discrimination, and clinical utility upon validation. By setting a threshold, the model can classify breast masses into low and high malignancy risk groups. Breast masses classified as low-risk can safely omitted from biopsy, thereby increasing the PPV for the remaining cases. As a result, this model improves the clinical utility of supplemental ultrasound in women with dense breasts.